The European Center for Medium-range Weather Forecasts' (ECMWF) archives of analyses and forecasts have been used extensively by the scientific community. In 1996, the Center completed a 15-year (1979-1993) reanalysis of data using a spectral version (T106 resolution, with 31 vertical hybrid levels or 17 pressure levels) of its operational data assimilation system and model. The motivation for the project was research applications in areas such as modelling, general circulation diagnostics and observing system performance. The principal components of the reanalysis are one-dimensional variational analysis of the cloud cleared radiances, optimal interpolation analysis and a global atmospheric model (Gibson et al. 1997). Sources of observational data for the reanalysis are data acquired in real time from the World Meteorological Organization's Global Telecommunications System (GTS); an archive of First GARP Global Experiment (FGGE) level II-b data; radiance data from TOVS [Television and Infrared Operational Satellite (TIROS) Operational Vertical Sounder]; cloud track wind data, generated for the FGGE level II-b dataset; ship and buoy observations from the Comprehensive Ocean Atmosphere Data Set (COADS); level II-b data from the Alpine Experiment; and portions of national archives from Japan and Australia. The Japanese archive provides additional data over the Northern Pacific and radiosonde data from around Japan. The National Meteorological Center at Melbourne provides surface pressures from pseudo observations called "PAOBS."

For this research, the ECMWF reanalysis data set was extracted from the archives at the National Center for Atmospheric Research (NCAR), Scientific Computing Division. Variables were obtained twice a day (00Z and 12Z), with 11 levels chosen for the upper-air fields, and a single level for supplementary fields on a 2.5 degree x 2.5 degree grid. The upper-air meteorological fields are geopotential (gz), temperature (K), zonal wind component (u), meridional wind component (v) and specific humidity (q) at levels (from bottom to top) 1000, 925, 850, 775, 700, 600, 500, 300, 250, 200 and 100 mb. Supplementary fields have also been obtained. They are large-scale precipitation, convective precipitation, surface flux of latent heat (to obtain evaporation) and top of atmosphere thermal radiation (to examine OLR). The upper-air fields consist of a 144 X 49 grid array, with each parameter available at each level. The supplementary fields are on a 144 X 73 grid array.

Cloud data from 1985 to 1991 from the International Satellite Cloud Climatology Project (ISCCP) are used for comparison and analysis in Chapters 4 and 5. Established as part of the World Climate Research Program, ISCCP is a cloud climatology that uses data from polar-orbiting and geostationary satellites. Two common types of data are used: high resolution images and sounder measurements. The high resolution images utilize the solar (VIS) and infrared (IR) spectra to observe changes in clouds and to measure surface properties. Sounder measurements use the thermal infrared and microwaves to measure temperature structure and humidity distribution. The data used for this research are pentad averages as prepared by Johnson (1996) from the 3-h or C1 data set distributed by the Langley Distributed Active Archive Center (DAAC).

The ISCCP cloud analysis (Rossow et al. 1991) is divided into three steps. (1) Cloud detection using VIS and IR radiances to make estimates of VIS and IR clear-sky background radiances. The degree of cloudiness at each place and time is determined by comparing the data to estimated clear-sky backgrounds. (2) Radiation analysis compares the data to a radiative transfer model to derive cloud optical depth (for the VIS spectrum) and cloud top temperature (VIS and IR spectrums). Cloud top temperatures are compared to TOVS profile data to estimate cloud top heights. (3) Through statistical analysis, monthly summaries of cloud-coverage, optical-depth and cloud-top-pressure are derived at 3-hourly intervals at a resolution of approximately 250 km. This reduces the volume of the data set. At this level in the processing, the 3-h ISCCP/C1 dataset becomes complete.

A data assimilation system interpolates the vast array of information from conventional and remote platforms to produce a grid of atmospheric values at regular intervals in space and time. This is done with a numerical model to produce a first guess and interpolation of available data which falls within a certain range of the first guess. Observations of evaporation, precipitation and radiation fluxes are not assimilated, and the gridded fields are modelled results obtained in the process of assimilating basic atmospheric variables. Therefore, their veracity might be doubted, so this chapter offers comparisons with observations and variables obtained from other assimilating systems for calibration purposes.

Fig. 2.1 compares average January (1985 to 1993) outgoing longwave radiation (OLR) values with those obtained from satellite estimates (Liebmann and Smith, 1996). The ECMWF produces higher values of OLR than observed in several regions: over the Pacific Intertropical Convergence Zones (ITCZ), in the region of the South Pacific Convergence Zone (SPCZ), from 140W to 100W, off the West Coast of South America, and from the Atlantic ITCZ to Africa. Over the South American tropics at about 15 degrees South, the monthly bias in Fig. 2.1 a shows a maximum (50 w/m2) in the difference between model and observed OLR, similar to the maximum (60 w/m2) daily variations (root-mean squared) in Fig. 2.1 b over the same region. The similarities between the monthly bias and the daily variations suggest that the monthly bias is the main contributor to differences between modelled and observed OLR. These differences may be attributed to inadequate representation of cloud tops or to the radiative transfer model. These biases are removed in the calculation of anomalies which are defined as differences from 15-year bi-weekly averages.

To examine the ECMWF precipitation representation, several comparisons have been made for the month of January (1988 to 1992) using various assimilated and observed precipitation datasets (Figs. 2.2-2.5). Convective and large-scale (dynamical) are the two precipitation types generated by the ECMWF model. Convection is tied to a simple cloud model (Gibson et al. 1997) for deep, shallow and mid-level convection. Large-scale models utilize bulk properties of clouds (Tiedtke 1993) and prognostic equations for cloud water/ice. Thus, precipitation and deep clouds are co-located in the analyses. Fig. 2.2 compares average January precipitation for the years 1988-1992 produced by the ECMWF to that produced by the Xie-Arkin (Xie and Arkin 1996) precipitation analysis. The difference between the two models (Fig. 2.2 c) shows the ECMWF producing considerably larger amounts of precipitation over the Andes corridor and over the Pacific ITCZs. Increased orographic convection in the ECMWF model can be attributed to the more realistic topographic representation given with the T106 spectral resolution, though the model erroneously places a large maximum in precipitation at around 32 S, 74 W. The The Xie-Arkin analysis gives a better representation of precipitation over the Andes than does the ECMWF. This is well illustrated when comparing it to results of other global precipitation observation projects such as those in Fig. 2.3.

Fig. 2.3 a shows the 1988-1992 January precipitation average for the Global Precipitation Climatology Project (GPCP), which uses a satellite gauge model (SGM) technique (Huffman et al. 1995). It combines precipitation estimates from microwave satellite data (86 GHz Goddard Scattering Algorithm 2), two infrared (IR) satellite datasets (Geo-IR and the NOAA 10 platform), rain gauges and numerical weather prediction models (the ECMWF operational T106 model). The merged precipitation dataset (Fig. 2.3 b ) combines monthly total precipitation data with oceanic Microwave Sounder Unit measurements at three channels (Spencer et al. 1993). Channel 1 is for cloud water and water vapor at 50.3 GHz, and channels 2 (53.74 GHz) and 3 (54.96 GHz) are for airmass temperature. The two data sets show accumulations near the southern tip of South America linked to midlatitude storm tracks. While the ECMWF assimilation shows this feature, the Xie-Arkin assimilation does not (see Fig. 2.2 c). Fig. 2.3 c shows pronounced differences between the observational datasets over the Pacific and Atlantic ITCZs with the merged analysis showing higher amounts. The maximum difference in precipitation over the Pacific ITCZ is smaller in the model-derived precipitation (Fig. 2.2 c) than it is with the satellite observation difference (Fig. 2.3 c). The model parametrizations are in closer agreement over the oceans.

Fig. 2.4 shows comparisons between two observational precipitation data sets (exact years used for the climatologies are unknown). The Jaeger data (Jaeger 1976) are from contemporary precipitation atlases over land. Over the ocean, Jaeger uses U.S. Marine Climatic Atlas data. The Legates-Willmott (Legates-Willmott 1990) data set applies bulk corrections to gauge values over land to correct for evaporation and wind catchment problems. Fig. 2.5 shows ECMWF-GPCP and ECMWF-Jaeger rainfall differences for January. Although the ECMWF consistently produces too much precipitation over the ITCZs and the SPCZ, the levels of uncertainty match the observational precipitation uncertainties in Fig. 2.2 c . Overall uncertainties in Fig. 2.5 are small in regions such as the SACZ and the subtropical plains of South America, the primary areas where precipitations patterns will be examined in this thesis.

Two comparisons are also made of upper-level variables. Fig. 2.6 compares the 200 mb U and V wind components of the ECMWF reanalysis and the NCEP reanalysis. Fig. 2.6 a shows the average ECMWF January and February 200 mb winds in 1993, while Fig. 2.6 b shows the vector difference between ECMWF and NCEP (the root-mean-square difference is contoured). The most pronounced differences appear over the Eastern Pacific from 120 W-80 W from the Equator to 30 S, where the root-mean square differences are between 8 to 10 m/s. Variation is also evident over the tropical Atlantic. Differences are small over South America and over midlatitudes. Fig. 2.7 compares ECMWF horizontal divergence to NCEP horizontal divergence at 200 mb. The ECMWF shows stronger divergence over the Pacific and Atlantic ITCZs, in the vicinity of the SPCZ (165 W, 10 S), and near the Equator at about 75 W. The NCEP analysis shows stronger divergence over Northeastern South America (2e-06s-1 more) at 40 W and 5 S.

Fig. 2.7 c and d compares 850 mb winds from the ECMWF and NCEP. Australian PAOBs were wrongly inserted in the NCEP assimilation over the southern oceans, which explains the differences observed between the two analyses south of 30 S over the oceans. Low level wind differences might result in substantial differences in moisture fluxes, an important quantity in this study that attempts to clarify the role of different moisture sources over South America. A comparison is presented in Fig. 2.8 between ECMWF and NCEP analyses for two events of interest. The overall patterns of moisture flux vectors are similar. Substantial differences, though, are shown in precipitating areas coinciding with moisture flux convergence (for example, over the SPCZ in Fig. 2.8 a and b). The ECMWF produces stronger moisture flux convergence over South America and in the SPCZ. This indicates that the differences are not only due to low level wind differences but are also linked to moisture and precipitation analysis differences.

Nogues-Paegle and Mo (NPM97) identify alternating wet and dry events over South America using empirical orthogonal function analysis on OLRA total anomalies during austral summer months from 1974/75 to 1997/95. Intraseasonal (10 to 90 days) filtered OLR anomalies are projected onto REOF 5 and times series principal components (PC's) are obtained. An event is labelled "positive" when the PC is greater than 1.2 of its own standard deviation and lasts for 5 or more days. Fig. 1.2 shows composite OLR anomalies of positive events (1979-1993) using observed (Liebmann and Smith 1996) and ECMWF assimilated data. Positive OLR anomalies, selected based on the specified REOF PC's are evident over the SACZ region. When the PC is less than 1.2 of its own standard deviation and lasts at least 5 days, it is labelled a "negative" event, Examples of negative events are shown in the composites of OLR anomalies in Fig. 1.2. An increase in convection in the SACZ occurs with negative REOF loadings (small OLR values over the SACZ), and a decrease in convective activity in the SACZ is found in the positive phase (higher OLR values over the SACZ, positive loadings in the REOF analysis). In this category, events are specified "with SPCZ." Figs. 1.1 a,b and 1.2 a,b show "with SPCZ" composites identified through the REOF analysis. In addition, Mo (personal communication, 1997) projects OLR anomalies onto REOF 5 without including a large portion of the Pacific Ocean (from 100 W to 360 W. The objective is to identify alternating wet and dry periods without regard to activity occurring in the region of the SPCZ. These events are categorized "SACZ only (see Figs. 1.1 c,d and 1.2 c,d )." For "with SPCZ," 19 positive and 14 negative events are identified from 1974/75 to 1994/95. For "SACZ only" cases, 19 positive and 16 negative events occurred. The events are listed in Table 4.1. This study examines the events during the austral summer months from 1979 to 1993, coinciding with the ECMWF reanalysis dataset. For the "with SPCZ" category, there are 15 positive events and 12 negative events during 14 summers. For the "SACZ only" category, there are 12 positive and 13 negative events during 14 summers.

The ECMWF re-analyzed fields from 1 November to 30 April were used to examine the events during the years 1979 to 1993. Daily values consist of an average of the 00Z and 12Z variables (helping to remove the diurnal cycle) for the upper-air data, while daily values for the supplementary fields are represented as an average of analyses/forecasts available at 00,06,12 and 18Z. A fourteen year climatology of the austral summer was produced to represent the seasonal cycle (1979-1992). The climatology is in the form of twelve two-week (approximately 15 days) averages beginning 1 November and lasting to 30 April. Composites of all events occurring between November 1979 and November 1993 were produced. In addition, variables for the days leading up to the event start date were taken into account. For each event, variables at day -8, day -6, day -4, day -2, and day -1 were computed to examine precursors leading to the beginning of the event. To examine the duration and demise of the events, variables were calculated at day 0, day +1, day +2, day +4, day +6, day +8, day +10, and day +12.

Ch.3 Climatology

Features of the atmospheric seasonal cycle as described by the ECMWF analyses from November through April are discussed in this section to 1) assess the veracity of the seasonal cycle in ECMWF analyses and 2) show the characteristics of the seasonally evolving mean climatology which is used to define anomalies.

Belts of high surface pressure in subtropical latitudes with low pressure towards the poles and the equator are well represented in the analyses. Oceanic surface pressure values are known to be lowest during winter and largest in summer, in contrast to what is observed over the continents (Fig. 3.1 ). The analyzed fields show higher values in the Pacific and Atlantic subtropical highs in November and April rather than in January. Extensive stratus decks over the cold currents of the eastern Pacific ocean shield the incoming solar radiation, modifying the resulting surface pressure and temperature. The surface analysis, therefore, may strongly reflect assimilating model physics.

Land areas respond to seasonal changes on the surface energy balance by developing low pressure during summer and monsoonal circulations which transport moisture laden air from the oceans. Regions with high orography channel this flow, forcing large scale ascent over the mountains and sustaining there high precipitation values. Fig. 3.1 shows surface wind divergence from the subtropical highs, with largest values over the eastern oceans. Strongest surface winds over the Pacific are found to the west of the South American coast, with a northwestward component which attains its maximum amplitude in boreal spring (Fig. 3.1c). This occurs at a time when the Pacific equatorial trough in these analyses is displaced towards the American coast.

During January (Fig 3.1b), the center of the Pacific trough is found at the dateline and displaced into the Southern Hemisphere (approximately 10 S). The surface wind at equatorial latitudes responds to the surface pressure gradient and attains a strong zonal component as it converges into the South Pacific convergence zone (defined below). The northeasterly flow into the east coast of South America stops at about 10 N during early November, and it moves southward to about 10 S as the season progresses (Fig. 3.1 b and c).

The regions of surface convergence are well reflected in the precipitation fields (Fig. 3.2 ). These are also co-located with warm sea surface temperatures, with well defined precipitation bands over the ITCZs, with monsoonal rains over Africa and South America, and with the Western Atlantic. Precipitation over the Western Atlantic is well-defined during the austral summer and exhibits a northwest-to-southeast tilt referred to as the South Pacific convergence zone (SPCZ). During this time of the year, precipitation over tropical South America attains maximum values.

The ECMWF re-analysis (with T106 resolution) is able to capture the double ITCZ symmetric about the equator during March to April (Fig. 3.2f). The validity of this feature has been confirmed with satellite estimates of rain using the Global Precipitation Climatology Project data set (Huffman et al. 1995). Also, the position of the Pacific trough in April (Fig. 3.1c) corresponds with this pattern. The ITCZ remains at about 10 N in all panels of Fig. 3.2, with precipitation attaining maximum values early in the season (Fig. 3.2a) when activity in the SPCZ region is weak. In contrast, as the SPCZ strengthens (Fig. 3.2c and d), the ITCZ weakens. A similar signature appears in the difference between evaporation and precipitation shown in Fig. 3.3 , with large negative values in the ITCZ early in the season (11/1-11/15) and large SPCZ values later in mid-summer (Fig. 3.3c and d). Over the tropical continents there is an excess of precipitation over evaporation during the 6 month period considered here. These precipitation patterns are sustained by convergence of vertically integrated moisture flux, diverging from subtropical latitudes where evaporation exceeds precipitation. In Chapter 4, we show similar oscillations between the ITCZ and the SPCZ with enhanced precipitation over tropical South America in much shorter time scales.

Over tropical South America, the Amazon basin is the center of maximum convection during December, January and February (DJF), where the central and southern portions receive more than 60 percent of their annual mean precipitation (Legates and Wilmott 1990). Convective activity penetrates farthest south in area and extent during this time and is mostly in the form of convective rain with a strong diurnal pattern (Silva Dias et al. 1987). An area of low-level moisture convergence extending from Central Brazil into the Atlantic (Fig. 3.2) also becomes strongest and most persistent during this time. This area is known as the South Atlantic convergence zone (SACZ). Fig. 3.6 shows the seasonal march of mean precipitable water averaged between 40W and 60W. The increase is noticeable in the Amazon basin and at subtropical latitudes. The precipitable water also depicts the SACZ in January (Fig. 3.6b). These are realistic features of the ECMWF re-analysis.

Seasonal changes over the Americas are strongly influenced by coastal geometry and orography. Similar to other monsoonal circulations, a well defined high pressure system is found at upper levels while the lower levels are modified by the prevailing orography. These and other features of the seasonal march over the Americas have been discussed by Horel et. al. 1989. The bi-weekly average march of 100mb wind along 20S is shown in Fig. 3.4a . The orientation of the vectors shows the seasonal westward progression of the anticyclone (referred to as the Bolivian High) and its gradual shift eastward late in the summer. A downstream trough is evident over Northeast Brazil, and zonal flow prevails south of 35S (Fig. 3.4b). These are dominant features during DJF which are well described by this re-analysis.

The seasonal march at 850mb is shown in Fig. 3.5a . Along 70W, the Andes act as a node, blocking low-level southerly flow to the West. Southeasterly flow west of the Andes appears independent of the seasonal march, suggesting the Pacific subtropical high (Fig. 3.5b) fluctuates little during the Summer. A broad area of northerly low-level flow throughout the Summer is also evident along the eastern slopes of the Andes. Over northern South America, northeasterly surface winds are predominant (Fig. 3.1 ), and strongest in April. At 850mb, flow with a more easterly component protrudes into the Amazon and turns southward as it approaches and is blocked by the Andes. This agrees with the results of Gandu and Geisler (1991), who simulated this blocked flow using a nonlinear, primitive equations model. The southward intrusion of moist oceanic air maximizes in January at about 40W, resulting from the westward-migrating Atlantic subtropical high. This is also reflected at the 1000mb level (Fig. 3.1b) where diverging surface winds move over the continent. Part of this research is to examine conditions in which moisture is intraseasonally transported to the subtropical plains by a low-level jet (NPM97).

The oceans play an important role in the flux of moisture. Large values of evaporation provide a source of moisture flux to regions with enhanced precipitation. Over South America, evaporation from the oceans exceeds precipitation throughout the Summer (Figs. 3.2,3.3). Along the coasts and over the Atlantic ITCZ, values are largest during February and March. Maximum values exceed 6 mm/day off the coast of Brazil at 15S and 7 mm/day over the Atlantic along 10N (Fig. 3.7 ) at a time when precipitation values over South America are most pronounced. Fig. 3.7 also shows strong values of evaporation over the Central and Western Pacific during January and February, coinciding with an enhanced SPCZ. Evaporation from the oceans is, therefore, a major source of vertically integrated moisture flux and contributes to evapotranspiration over the continent. Anomalies from the seasonal march are presented in Chapter 4 for cases of enhanced precipitation over tropical South America.

Ch.4 Negative Events

Periods of enhanced convection over the South Atlantic Convergence Zone have a pronounced effect not only over tropical South America but also in neighboring regions, where compensating sinking motions promote suppressed precipitation. This chapter documents the large scale circulations associated with intensification of convection in the SACZ and examines the difference in local circulations resulting from different convective regimes in the Pacific Ocean. It is shown that strong SACZ events develop in about 10 days, with two different convective regimes over the Pacific Ocean, as shown in Fig. 1.1 . Events with suppressed convection over the SPCZ (150W 20S) and enhanced convection over Australia and the SACZ are components of a large scale system which evolves from its opposite phase in about 10 days. In contrast, there is a second type of evolution of enhanced convection in the SACZ with roots over South America and the Pacific Ocean, without any apparent ties to the Pacific Ocean. Such cases are accompanied by an enhanced SPCZ.

The chapter is divided into five subsections as follows: first, OLR and precipitation fields are discussed; followed by a description of associated upper-level circulation anomalies; and a discussion on the possible role of synoptic scale transients. The fourth subsection examines components of the atmospheric hydrological cycle for these events. Substantiation of results obtained from the ECMWF reanalysis is then presented by comparing ECMWF analysis composites obtained for events over a six year period with cloud coverage obtained from the ISCCP and OLR observations.

Rotated empirical orthogonal function (REOF) analyses of OLR anomalies, described in section 2.4, are used to identify periods of alternating wet and dry conditions over subtropical South America during summer. Two types of REOF analyses are conducted. The first focuses on South America, extending from 40 S to 10N and from 100 W to the Greenwich Meridian (denoted as REOF A) while the second is applied to a sector from 180 W to 20 E and from 40 S to 40 N (denoted as REOF B). Filtered data is projected onto these EOFs to obtain time series of principal components. An event is chosen when the magnitude of the PC is greater than 1.2 of its own standard deviation and lasts for 5 or more days. Events selected by REOF A and B are used to study SACZ cases with possible connections to atmospheric processes over the Pacific Ocean. These are referred to as "SACZ with SPCZ (abbreviated "with SPCZ")." Events chosen with REOF A but not with REOF B are referred to as "SACZ only." The negative events are listed in Table 4.1 (below) and are shown to have an average duration (as defined in Chapter 2) of about a week. Events which occur during warm ENSO years are denoted by an asterisk. The analyses reveal that the "with SPCZ" cases are more than twice as likely (64% versus 25%) to occur during warm ENSO years. NPM97 found that these events were 3 times more likely to occur during warm and cold ENSO years for the period 1973-1993.

Table 4.1

w/SPCZ Negative	SACZ Negative
01/20-02/02 '77*	02/04-02/09 '75
01/23-01/28 '82	03/24-03/28 '76
03/03-03/07 '83*	01/14-01/23 '81
01/22-01/29 '84*	03/25-03/30 '81
03/08-03/12 '84*	03/08-03/12 '82
11/24-12/01 '85	01/17-01/30 '85
01/26-01/31 '87*	02/27-03/09 '85
02/15-02/20 '87*	11/03-11/08 '85
01/07-01/11 '88	12/23-12/28 '85
03/11-03/15 '89*	02/16-02/20 '86
03/22-03/30 '91	12/25-12/31 '86*
01/16-01/25 '92*	02/24-02/28 '90
03/26-03/31 '93*	11/11-11/16 '91*
03/03-03/09 '94	11/02-11/07 '92*
-	03/05-03/13 '93*
-	02/02-02/12 '95
Total=14	Total=16
Ave Time=7.1 dy	Ave Time=7.4 dy
64% Warm ENSO	25% Warm ENSO

* denotes events which occurred during warm ENSO years

Clarification of the relationship between ENSO events and the "SACZ with SPCZ" cases is further explored by comparing OLRA composites obtained from the intraseasonal filter (IS) (discussed in Chapter 2) with those obtained from a low pass filter (LP) that removes time scales less than 10 days. A third type of composite is also produced. It consists of events in the list for REOF B but not in A. These are referred to as "SPCZ only". All cases are shown in Fig. 4.1. Comparison of a) with d) and b) with e) shows remarkable similarity between the IS and LP filters, indicating little contribution to this pattern from interannual variations. Explanation for the "with SPCZ" pattern to appear during ENSO events is more readily seen comparing Figs. 4.1 c and f. This shows an amplified OLR anomaly response west of South America for the interannual time scale compared to what is observed on the intraseasonal scale. ENSO events favour eastward displacement of the SPCZ, a feature associated with some cases of enhanced SACZ (see Fig 4.1b). Nevertheless, ENSO events do not have a signal on the SACZ, as shown by the similarity over the SACZ region between Fig. 4.1 c and f. During the average lifetime of a typical ENSO episode, the number of cases with enhanced and suppressed convection over the SACZ will average out any ENSO signal.

Support for this weakening of the ENSO signal is apparent in Fig. 4.2 a through d, which displays the evolution of the outgoing longwave radiation anomalies. It compares well with the "seesaw" pattern NPM97 identified using observed OLR anomalies obtained from National Environmental Satellite, Data, and Information System averages of individual satellite scans. They linked the evolution of the "seesaw" pattern to convection over the Pacific Ocean. Similar to the results of NPM97, Fig. 4.1 c and d show that convection is supressed in the region of the SPCZ, while an enhancement in tropical convection occurs in the Central Pacific. Over Eastern Australia, similar sign reversals are seen in the OLR anomalies, evolving towards a regime of enhanced precipitation over Central Australia and suppressed convection towards the north. The suppressed convection over the Maritime Continent is apparent in the ENSO mode, and this also tends to favour the occurrence of the "with SPCZ" events during ENSO years. However, the signal in the interannual time scale is damped by intraseasonal variations.

The "SACZ only" composites in Fig. 4.2 (e-h) reveal a stationary pattern over South America (no "seesaw"), as the convective area over the SACZ builds with time. Signatures of low OLR anomalies are already present over the SACZ at day -8. The ITCZ weakens with time over Central America and the Gulf of Mexico as low OLR signatures shift northward. Over the SPCZ, activity is not suppressed as it is for the "with SPCZ" cases. In contrast to the evolving amplification in low OLR over Australia, a more stationary, weaker area of low OLR anomalies over that region develops for the "SACZ only" cases by day +1. Therefore, the mechanisms of formation in the "SACZ only" cases appear to differ significantly from those of the "with SPCZ" cases.

Analyses of OLRA have been offered as proxies of convection. A more realistic analysis follows, showing the precipitation fields derived by the assimilation system during the course of data insertion.

The OLR anomalies in Fig. 4.2 indeed correspond to enhanced precipitation patterns, as shown by the composites in Fig. 4.3 . The "with SPCZ" composite (Figure 4.3a) shows an enhanced SACZ with precipitation extending in an elongated arc into the Southwest Atlantic. The Pacific ITCZ is well-defined by an extensive area of precipitation, and a northwest-to-southeast-oriented band of precipitation anomalies is evident over Australia. 200 mb horizontal convergence and divergence correspond well with areas of suppressed and enhanced precipitation, respectively. Over the SACZ, horizontal divergence values of up to 3 X 10-6 s-1 occur (Fig. 4.3a), while convergence values are high in the region of suppressed convection over the subtropical plains of South America. As shown in Fig. 4.2, high OLR anomalies strengthen with time over the subtropical plains as precipitation over the SACZ strengthens ("seesaw" pattern). In the region of the SPCZ, precipitation is suppressed and has been displaced eastward, corresponding to the OLR in Figure 4.2d.

For the "SACZ only" composite (Fig. 4.3b), precipitation is more continental in nature and stronger, extending uninterrupted into Northeast Brazil and with less protrusion into the Southwest Atlantic. Maximum anomalies exceeding 7 mm/day are found in the vicinity of 20S, 45W. Surrounding this intense convection are extensive areas of negative precipitation anomalies. Compensating subsidence around the active SACZ appears more widespread for the "SACZ only" cases, though negative precipitation anomalies are generally less intense over the subtropical plains of South America. Precipitation is in an opposite phase over the Pacific at the Equator compared to the "with SPCZ" cases. The Pacific ITCZ is suppressed over the equator, and 200mb horizontal divergence and precipitation extend farther northward. The SPCZ is also at a normal position over the Western Pacific (Fig. 4.3b). Over the Eastern Pacific, off the coast of Central America, low OLR anomalies (Fig. 4.2d) show evidence of cirrus blowoff (not precipitation) from convection to the south, as precipitation anomalies are not shown in region. From this discussion of the two types of negative cases, differences in large scale organized convection for the two types of SACZ events are apparent. An analysis of the upper-level flow patterns provides further insight.

To diagnose the upper-level circulations during negative events, composites have been compiled for the 200mb and 500mb levels. Fig. 4.4 shows precipitation and 200mb height anomalies for each case and the difference (Fig. 4.4c) between the two types of enhanced SACZ. Four well-defined northwest-to-southeast-oriented precipitation bands are evident (Fig. 4.4a from the SACZ eastward into the Central Pacific, extending from the tropics into midlatitudes. These bands acquire a more zonally-oriented tilt as they merge into the westerly flow. Approximately 5 waves are found in the 200 mb height anomaly pattern between 40 and 50S, suggesting a possible linkage between precipitation and height fields. This is confirmed in Fig. 4.4d, which shows precipitation anomalies south of 20S co-located with regions of negative height anomalies at 1000 mb. The 200 mb negative height anomalies are displaced toward the west (or southwest) of the surface anomalies, as expected for baroclinic waves. In contrast, waves for the "SACZ only" cases are longer (wave 2) and weaker. A wave 5 pattern is seen in the differences (Fig. 4.4c) with values generally higher than those of either Fig 4a or b. This indicates the wave patterns are mainly out of phase. An exception to this is the negative low height anomaly located over the South Pacific off the Southwest Coast of South America, which appears in both cases. The 1000 mb height field exhibits a more complex wave pattern than at 200 mb for the "SACZ only" composite (Fig. 4.4e). The largest amplitudes are found at 50S in the vicinity of South America.

Another characteristic of the upper-level flow field is the more active wave pattern over North America for the "SACZ only" composite (Fig. 4.4b), showing high height anomalies over the Western U.S. and low height anomalies over the Eastern U.S. Convection over the Central and Eastern Pacific is shifted northward, possibly contributing to the amplification over North America. The possible association of the "SACZ only" negative patterns with the amplification over North America will be further examined in section 4.4.

Results presented in this section have thus linked the tropical convergence zones to a baroclinic wave number 5 observed in the height anomalies for the "with SPCZ" cases. This is also substantiated by the vertical structure of wave height anomalies at 50 S, shown in Fig 4.5a. In general, height anomalies maximize between 250 and 300 mb with large central values observed in all waves. In contrast, the wave 2 pattern of the "SACZ only" cases is weaker and appears to propagate upward eastward of 60 E. Fig 4.5c shows the structure of the stationary waves at 50S. These are characterized by a strong wave number 1, which has been identified (e.g., Hoskins and Karoly 1981) with a meridionally propagating Rossby wave generated by Antarctica. Both cases show deviations from the average pattern, indicating that the anomalies are not a result of an amplification of the climatological state. Therefore, the next question to address refers to the event evolution, in an effort to quantify the degree of transience associated with each type of event.

The strength of the synoptic eddies is quantified by the variance of 200 mb heights within each event, and then the square root is taken from the mean variance for all events. Fig. 4.6a shows this quantity for the "with SPCZ" cases. The Southern Hemisphere appears more variant than the Northern Hemisphere. When the "SACZ only" composite is subtracted from the "with SPCZ" composite (Fig. 4.6b), the latter contributes more strongly to transient activity at midlatitudes. For example, high variability between 0 and 30 W may be a reflection of more active synoptic patterns over the Southwestern Atlantic. In the average, this might support the extensive, arc-like nature of the precipitation over this region (see Fig. 4.3a). Similar associations may be established for other convergence zones discussed previously which are apparent for the "with SPCZ" cases.

The previous discussion relates to the synoptic time variability contained within individual events. Here, the averaged composite evolution is shown with a hovmoller diagram of the 200mb height anomalies along 50S for "with SPCZ" cases (Fig. 4.7). An amplified (wave 5) pattern is shown, specifically during the onset (day 0) of the events. Southern South America is influenced by a broad area of high height anomalies which maximize at day 0. Near 120E and south of Australia, strong negative anomalies reach a maximum by day +6. These anomalies are also visible in the horizontal composite (Fig. 4.4a).

The more stationary nature of the "SACZ only" cases is seen in the composite evolution of the 500 mb height anomaly shown in Fig. 4.8. The wave pattern over the southern Pacific Ocean is already evident 8 to 6 days prior to the onset. Largest values are found in association with the negative anomalies between 60W and 120W (Fig.4.8d). As evident in Fig. 4.6b, the region between 60W and 120W is also less transient than are the "with SPCZ" cases. Whether the precursors to the height anomalies in the 0 to 120 W region are found over the Pacific or are associated with the strengthened SACZ pattern (Fig. 4.4e) is not clear at this time. An examination of the hydrological cycle for the negative cases follows.

This section examines the moisture sources contributing to the two cases in which the SACZ is enhanced. The differences in precipitation are linked to different moisture sources. Fig. 4.9 shows the 850mb wind field and precipitation anomalies for "with SPCZ" and the difference between the two types of negative cases considered. The precipitation differences show that precipitation is more intense over tropical South America for the "SACZ only" cases, but it is more intense over the Pacific ITCZ for the "with SPCZ" cases. Features of the low-level circulation include subtropical highs located on the east and west sides of South America, zonal westerly winds at midlatitudes, and a strong easterly flow of tradewinds into the interior of Northern South America.

In examining the differences between "with SPCZ" and "SACZ only", several features are apparent. The Pacific subtropical high appears stronger for the "SACZ only" cases Fig. 4.9b). The cool southerly flow over the west coast of South America is much different from the warm, moist northerly flow associated with the subtropical high on the east coast. Over the Southwestern Atlantic, the westerly return flow in the region of precipitation anomalies is stronger for the "with SPCZ" cases. At around 40W, 30S, Fig. 4.9b shows flow for the "SACZ only" cases as more meridional closer to the coast, producing an anticyclone in the differences. This supports the more continental nature of the precipitation for the "SACZ only" events. The orientation of the Atlantic subtropical high differs between the two cases. For "with SPCZ," the ridge axis is northwest-to-southeast, while for "SACZ only", the ridge axis is more zonal. This is consistent with the general northwest to southeast prevailing orientation found in the convergence zones for the "with SPCZ" cases.

Flow regimes also differ over the Equatorial Pacific. The "with SPCZ" composites show a stronger easterly component over the Pacific ITCZ. For the "SACZ only" cases, easterly flow is not as intense but remains zonal for a longer distance. Low-level flow turns toward the south at about 180 degrees longitude for the "with SPCZ" cases, in an area of precipitation anomalies between the Equator and 10S. Farther south, in the region of the SPCZ, the easterly flow has a stronger meridional component for the "SACZ only" cases. In the Northern Hemisphere, at about 170W, 20N, meridional components also appear in the difference between the cases (Fig. 4.9b), suggesting a stronger meridional component to the flow for "SACZ only" cases. This is also a region where precipitation anomalies have shifted northward.

In section 4.2, the upper-level height anomalies for the "SACZ only" cases (Fig. 4.4) showed an amplification in the flow over North America, with lower height anomalies over the Eastern U.S. 850 mb composite temperatures and precipitation anomalies (used as a reference) in Fig. 4.10 also show the influence from the "SACZ only" cases. Between 20N and 30N, a two degree temperature anomaly (Fig. 4.10c) appears over the Gulf of Mexico, showing cooler temperatures for the "SACZ only" cases. Such cold anomalies may be related to cold outbreaks from the winter hemisphere, and they sustain the anomalous precipitation pattern seen in the Eastern Pacific Ocean between 90 and 105W (Fig. 4.10e). In contrast, the low level winds for the "with SPCZ" cases are very different in the vicinity of North America. Southerly flow is increased (Fig. 4.10d), associated with low anomalies of the height field over the continental USA and Mexico. Low level wind and precipitation anomalies over equatorial latitudes over Northeastern Brazil suggest that the continental precipitation anomaly is supported by southerly continental air masses (fig 4.10d) while the displacement over the oceans results from northerly flow (fig 4.10.e). These relationships are further explored by considering the role evaporation and vertically integrated moisture fluxes play in maintaining precipitation anomalies.

Fig. 4.11 shows evaporation and precipitation composite anomalies for the two cases. For both cases, the region of the SACZ is characterized by negative evaporation anomalies, while regions of suppressed convection (over the SPCZ in Fig. 4.11a, for example) tend to have positive evaporation anomalies. This suggests a lack of feedback between precipitation anomalies and the underlying surface fluxes. Whether this is true for the time scales of the anomalies considered here or an artifact of the assimilating model physics remains to be determined. Overall, the evaporation anomalies shown in the ECMWF analysis are small, at least an order of magnitude smaller than the precipitation anomalies. This suggests that evaporation has a small role as a moisture source for these events, and therefore, precipitation anomalies must be mostly maintained by convergence of moisture flux.

In Fig. 4.12 a and b , moisture flux convergence is co-located with precipitation anomalies (seen in Fig. 4.3). The extensive arc of convection defining the SACZ for the "with SPCZ" cases is apparent in the convergence of moisture flux, which extends southeastward to about 15W. In the ITCZ over the Eastern Pacific, flux convergence is strong for the "with SPCZ" composites (where precipitation anomalies are located), while it is absent for the "SACZ only" cases. Over the West Pacific, convergence of moisture flux is well-defined over the SPCZ and north of the Equator for the "SACZ only" cases. Similarly, for "with SPCZ" cases, flux convergence coincides with precipitation anomalies. Off the west coast of South America, moisture transport is weaker for the "with SPCZ" than for the "SACZ only" cases, in contrast to the strong moisture flux on the eastern side of the continent. The moisture flux vectors appear to follow the low-level winds, suggesting that the main contributor to the convergence of moisture flux is given by the wind convergence multiplied by the specific humidity rather than moisture advection.

Differences in moisture transport between the two cases are illustrated in Fig. 4.12c. Moisture transport differences are large, as maximum arrows are greater than 50 percent of the original values. At midlatitudes, the reflection of higher wavenumbers for the "with SPCZ" cases is shown in the vector wave pattern. Over South America, the continental nature of the anomalous convection for the "SACZ only" cases is shown (as flux divergence), similar to the precipitation anomalies. Over the Southwest Atlantic, the "with SPCZ" cases show significant moisture flux convergence (for example, near 35W, 40S). The zonal component of moisture flux dominates convergence over the Pacific ITCZ. Oceanic moisture transport is stronger in this region for the "with SPCZ" events. Over the SPCZ and in the Northern Hemisphere (around 180W), larger areas of moisture convergence from the "SACZ only" cases are shown as flux divergence. Over Australia, the precipitation anomalies observed for the "with SPCZ" events are accompanied by strong moisture flux (110E). These composites show that the major moisture source for the anomalous precipitation events may be associated with anomalous moisture flux convergence.

A description of two different patterns of enhanced convection over the SACZ has been given to this point based on the ECMWF reanalysis. Some of the described quantities, such as precipitation and evaporation, are derived by the assimilation system and may deviate from reality depending on the accuracy of the physical parameterizations of the assimilating model. Next, data obtained from satellite measurements are used to ascertain, to the extent possible, the veracity of the convective patterns discussed above.

To look at the reliability of the ECMWF assimilated data, composites of total cloud and OLR fields have been assembled from the ISCCP dataset. The composites are for negative events which occurred from January 1985 to March 1991. The ECMWF reanalysis global precipitation field has been used to make several conclusions, but one way to verify the validity of the data set is to compare it to observed cloud data. This is possible since the ECMWF analysis sytem carries prognostic equations for cloud water and cloud cover (Tiedke 1993), with cloud water/ice losses resulting from cloud evaporation and precipitation processes. The approach in this section is 1) to compare precipitation patterns with cloud coverage derived from satellite measurements, and 2) to compare ECMWF derived OLR fields with those obtained from satellites. Approach 1) is done to offer some validation of the ECMWF. Approach 2) involves radiation calculations as well as cloud processes, and it gives an overall assessment of the veracity of both parameterizations.

A comparison of Fig. 4.13with Fig. 4.15 shows the precipitation and OLR differences from the ECMWF, as well as deviations from observed OLR (Fig. 4.13c), to verify differences between the two SACZ event types previously discussed. The model-derived precipitation and OLR fields tend to correspond well, with low OLR values and high precipitation estimates over the Southern Hemisphere convergence zones (FIG. 4.13). Differences exhibit consistent signals, though it is apparent that the ECMWF OLR is about 20 to 40 w/m2 too high over regions of convection (Fig. 4.13 c). Over South America, for example, the "with SPCZ" cases have lower precipitation values and higher OLR values. Fig. 4.15c shows the difference in observed OLR (from Liebmann and Smith, '96). The ECMWF differences (Fig. 4.15b) tend to match observations over South America and over the SPCZ. Corresponding to the observations, the SACZ extends over the Southwest Atlantic for the "with SPCZ" events, and it remains more confined to the continent for "SACZ only" events. The differences over Australia are similar for both the assimilated and observed data. Over the Equatorial Pacific, the ECMWF differences appear too large compared with observations, but over the Northern Hemisphere, the observations pick up on the displaced OLR for the "SACZ only" cases.

Composites of ISCCP cloud data for negative events are shown in Fig. 4.14 and 4.16. They show cloud percentages for high, middle and low clouds (prepared by Johnson, 1996). The main feature of the low cloud differences is the large area of low cloudiness off the West Coast of South America, which occurs predominantly for the "with SPCZ" cases. Cloud climatologies for the austral summer (eg. Johnson, '96) show pronounced low stratus decks in this region, but the broad area covered in Fig. 4.16c exceeds the climatological amount. Fig. 4.11shows pronounced negative evaporation anomalies over the Southeastern Pacific, especially for the "with SPCZ" cases. Extensive cloudiness (contributing to lower amounts of evaporation) over this region plays a role in the earth radiation budget, as solar radiation becomes less influential with the insulating layer of clouds. Due to the cool waters and predominant southerly flow, the low clouds in this region are generally layer clouds (stratus) which are associated with warm air over a cold surface (Wallace and Hobbs 1977). The stratus decks are more extensive and persistent for the "with SPCZ cases," consistent with the warmer air temperatures observed at low levels. Differences in temperature between "with SPCZ" and "SACZ only" at 80 W are positive in the lower half of the atmosphere, with values of 3 degrees found at 850 mb and 22.5 S (not shown). These warmer temperatures and concommitant higher static stability at low levels are consistent with the more extensive stratus decks off the Chilean coast.

Results presented in this section confirm the adequacy of the ECMWF assimilated data to describe atmospheric patterns related to the two different classes of SACZ enhancement. The primary source of remaining uncertainty is with the evaporation field and whether the lack of local feedback between evaporation and precipitation is a realistic feature of the atmospheric component of the hydrological cycle over time scales of 1 to 2 weeks, which are of interest to this current research.

Ch.5 Positive Events

Rainfall patterns over the subtropical plains of South America have important agricultural and economic factors during the austral summer. This chapter focuses on periods of suppressed convection over the SACZ in which precipitation is abundant over the plains. As shown in Chapter 4, an intensified SACZ results in compensating subsidence that locally suppresses convection in surrounding regions. It is shown that increased precipitation over the plains occurs at the other extreme, when the SACZ weakens, on a similar timescale of about 10 days. This chapter examines two cases of suppressed convection over the SACZ (shown in Fig. 1.2). One type of event evolves with ties to the larger scale system, representing the opposite phase of the "seesaw" described in Chapter 4 (Fig. 1.2 a). The other regime originates more locally and with more intense local characteristics (Fig 1.2 b).

The chapter is divided into five subsections as follows: OLR and precipitation fields are discussed; a description of associated upper-level circulation anomalies follows; third, a discussion on the possible role of synoptic-scale transients; next, components of the atmospheric hydrological cycle are examined; and finally, event composites from the ECMWF reanalysis for a six-year period are compared with cloud coverage obtained from the ISCCP and OLR observations.

The "with SPCZ" and "SACZ only" cases (see section 4.1) are identified by REOF analyses of OLR anomalies (see section 2.4), as in Chapter 4. Table 5.1 (below) lists the positive events, and they are shown to have an average duration of about a week. Events which occur during warm ENSO years are denoted by an asterisk. Although not nearly as strong as for the negative events, the positive "with SPCZ" events are more closely associated with warm ENSO events than are the "SACZ only" cases (37% versus 21%). The "with SPCZ" positive events are one phase of the "seesaw" from enhanced to suppressed convection over the SACZ (the positive signal has been shown to occur 8 days from onset of a negative event). Results in Chapter 4 show that the opposite phase (negative "with SPCZ" composite) has a much stronger ENSO signal. To further examine these relationships, OLRA composites from the intraseasonal filter (IS) are compared to those obtained from a low pass filter (LP) that removes time scales less than 10 days. The "SPCZ only" composite is also produced (see section 4.1). The cases are shown in Fig. 5.1. Comparison of a) with d) and b) with e) shows that the interannual variations are more pronounced, especially over the Pacific Ocean. This indicates strong contributions to these events from interannual variations. For the "with SPCZ" cases (b and e), the lack of an ENSO signal over the SACZ is supported by the intraseasonal "seesaw," which tends to average out any signal. Over the subtropical plains, low OLR anomalies indicate abundant precipitation south of the suppressed SACZ. This is a region Ropelewski and Halpert (1996) identify as exhibiting Southern-Oscillation related precipitation relationships. Enhanced precipitation in this region matches a warm ENSO signal. Comparison of c) with f) supports the higher correlation of positive "with SPCZ" events to warm ENSO events. Over the plains in Fig. 5.1c, OLR anomalies are weaker than those of 5.1f. Positive events for the intraseasonal time scales are not apparent when compared to the interannual signal (Fig. 5.1f), which also exhibits ENSO characteristics.

Table 5.1

w/SPCZ Positive	SACZ Positive
03/07-03/12 '76	02/15-02/19 '75
01/12-01/16 '77*	12/06-12/11 '75
02/09-02/14 '77	01/16-01/21 '76
12/20-12/25 '80	03/24-03/30 '80
02/15-02/19 '81	01/30-02/04 '81
02/13-02/23 '82	12/20-12/25 '81
02/16-02/26 '83*	11/01-11/08 '84
01/10-01/18 '84*	02/04-02/10 '85
02/13-02/22 '85	12/08-12/13 '85
11/12-11/20 '85	01/18-01/22 '86
01/05-01/10 '87*	11/18-11/25 '86*
02/23-03/03 '87*	03/06-03/13 '90
01/15-01/23 '88	11/22-11/26 '91*
03/26-03/31 '88	01/01-01/05 '93*
12/07-12/12 '88	11/11-11/16 '93*
01/25-02/02 '89	02/02-02/06 '94
02/16-02/23 '92*	11/05-11/10 '94
01/27-01/31 '93*	12/11-12/17 '94
02/23-03/02 '95	01/14-01/18 '95
Total=19	Total=19
Ave Time=7.6dy	Ave Time=6.2dy
37% Warm ENSO	21% Warm ENSO

* denotes events which occurred during warm ENSO years

Evolution of the positive "with SPCZ" cases (Fig. 5.2 a-d) shows a transition over the SACZ from low to high OLR anomalies. By 4 days prior to event onset, the SACZ becomes suppressed with anomalies up to 25 W/m2. By day 0 (Fig. 5.2 d), the signal is pronounced over the SACZ, while a weaker sign of increased convection over the subtropical plains is evident. At higher latitudes, OLR becomes a less reliable proxy of convection. (the composites in Fig. 1.2 b and d show a stronger signal over the subtropical plains). Analysis of the precipitation field will show that the convective activity over the plains is more enhanced than what may be inferred from Fig. 5.2 d. Similar to the evolving conditions over the SACZ, the low OLR anomalies over Australia at day -8 (reminiscent of the negative phase) weaken by day 0, and the SPCZ becomes more pronounced in its normal position by event onset.

The "SACZ only" events (Fig. 5.2 e-h) develop in a more stationary manner, similar to the negative cases (Fig. 4.2 e-h). Over the SACZ, high OLR values are evident at day -6 and slowly build with time, as do the low OLR anomalies over the subtropical plains to the south. High OLR anomalies are located over the Equatorial Pacific as convection is displaced northward into the Northern Hemisphere. Anomalies over South America are stronger than they are for the "with SPCZ" events. As with the negative cases, the mechanisms of formation for the two types of positive cases differ. A validation of the OLR anomalies as proxies of convection is shown with the following analysis of the precipitation fields.

Three bands of positive precipitation anomalies appear in the Southern Hemisphere for the "with SPCZ" events (Fig. 5.3 a). Precipitation is pronounced over the subtropical plains of South America, in the SPCZ and over Australia. Over Australia, precipitation is not as intense as it is for the negative phases of the "seesaw." OLR anomalies in Fig. 5.2 d a correspond well with the precipitation. The region over the SACZ is well-defined (Fig. 5.3a) with negative precipitation anomalies exceeding 5 mm/day, and high values of upper-level convergence. This composite shows the counterpart to the negative "with SPCZ" events. With roots to the large-scale circulation, precipitation and divergence anomalies do not show as significant a signal as do the more locally oriented "SACZ only" cases in Fig. 5.3b.

Precipitation anomalies for the "SACZ only" cases (Fig. 5.3b) are much stronger, particularly over the SPCZ and over South America (two main bands of precipitation extend from NW to SE in the Southern Hemisphere). They show more pronounced negative precipitation and upper-level convergence anomalies over the SACZ and an intense, more extensive band of precipitation anomalies to the south over the plains (precipitation anomalies are greater than 6 mm/day at about 50 W). In the Equatorial Eastern Pacific, a belt of enhanced precipitation is displaced northward to about 10 N. The pattern has characteristics of the warm ENSO phase. Therefore, as with the negative events (Ch. 4), the nature of the two types of positive cases differ. An analysis of composite upper-level circulations follows.

Composites of 200 mb and 1000 mb height and precipitation anomalies are shown in Fig. 5.4. An upper-level wave number 3 pattern is shown in the Southern Hemisphere at midlatitudes (Fig. 5.4a) with the most pronounced high at 100 W. Southeast of Africa, wave activity is suppressed and negative precipitation anomalies over Madagascar contrast with the anomalous precipitation band seen with the negative cases (see Fig. 4.4 a). This shows(as mentioned previously) how the "with SPCZ" positive cases are the o pposite phase of the "seesaw," so strong signals seen with the negative weaken in the positive phase. At 1000mb (Fig. 5.4 d), the wave pattern is strongly amplified near South America, and low height anomalies are tied to positive precipitation anomalies (for instance, over Australia). Surface ridges are in place over the SACZ, co-located with negative precipitation anomalies.

The wave pattern for the "SACZ only" cases (Fig. 5.4 b) shows a less well defined Southern Hemisphere wave number 3 in mid-latitudes. Near South America, the midlatitude 1000 mb height fields show a trough-ridge-trough pattern in Fig. 5.4 d, while the "SACZ only" events (Fig. 5.4 e) show a ridge and a very intense trough over the Southwest Atlantic. The height field is amplified (see Fig. 5.4 b and e) with very low anomalies in the height field at 30 W. Garreaud and Wallace (1997) found that summertime incursions of midlatitude air are predominant features of the South American circulation. They move equatorward at about 10 m/s and maintain identities for about 10 days. These incursions are preceded by an approaching upper-level trough which deepens the surface low over the central plains. The southward advection of moist air makes conditions favorable for deep convection over central and southern Argentina. The localized nature of the "SACZ only" events suggests that the passage of these strong frontal systems may play a role in the observed signature. The composites in Fig. 5.2 suggest a northward migration of OLR anomalies from mid-latitudes to the subtropics, moving at a much slower rate of about 10 degrees latitude in 10 days. Over the SPCZ, 200 mb negative height anomalies are displaced toward the west of the surface anomalies, as expected for baroclinic waves. The difference in height and precipitation anomalies (Fig. 5.4 c) is well defined and suggests that the patterns are somewhat out of phase. The higher magnitude of "SACZ only" precipitation anomalies in the Southern Hemisphere is also illustrated. Over the Northern Hemisphere, the wave pattern is much more active at midlatitudes for both cases than was found for the negative cases. At the 1000 mb surface over the Midwestern U.S., the "SACZ only" cases show positive precipitation anomalies and southerly flow over the Eastern U.S., as indicated by the position of the height anomalies (Fig. 5.4 e).

The upper-level analyses thus show that large-scale circulations for "with SPCZ" events are more active, while "SACZ only" events have more locally amplified features near South America. This can also be shown with cross sections of meridional flow anomalies at 60 W (Fig. 5.5). A low-level northerly jet which transports moisture southward is established parallel to the Andes at midlatitudes (this will be discussed in further detail in subsection 5.4), and northerlies aloft maximize at 35 S. The upper level northerlies are associated with the approaching trough off the west coast of South-America. Meridional anomalies for "SACZ only" events are stronger at low-levels and extend farther south aloft than those for "with SPCZ" events, with values up to 6 m/s at 300 mb (at 35 S, Fig. 5.5 b). The upper level flow implies substantial differences from climatology (Fig. 5.5 c), which depicts prevailing southerlies at upper levels associated with an upper level anticyclone (Fig. 3.4). The next section looks at transient activity and the evolution of the events.

As in Chapter 4.3, the strength of the synoptic eddies is quantified by the variance of 200 mb heights within each event, and then the square root is taken from the mean variance for all events. This quantity is shown in Fig. 5.6 a for the "with SPCZ" cases. The Northern Hemisphere and Southern Hemisphere both appear considerably variant at midlatitudes. In the Southern Hemisphere, when the "SACZ only" composite is subtracted from the "with SPCZ" composite (Fig. 5.6b) both cases tend to contribute to transient activity at midlatitudes, though the "with SPCZ" cases are predominant. While the "with SPCZ" cases are more transient over the Pacific (as in the negative cases in Fig. 4.6), the SACZ cases show transient activity over the Southwest Atlantic (as noted by the -20 m difference). This supports conclusions in section 5.2 that strong cold air outbreaks may play a role in the signal over South America for the "SACZ only" events. Over North America, transient activity is much more pronounced than was found for the negative events.

The averaged composite evolution is now shown with hovmoller diagrams of the 200 mb height anomalies along 50 S for both types of cases (Figs. 5.7 and 5.8). For the "with SPCZ" composite, a wave pattern is in place from 180 W eastward by day -4. In support of the "seesaw" concept, this is a similar type of amplification observed for the negative cases at day 0 (see Fig. 4.7). Near Indonesia (from 120 E to 180), the strengthening of the high height anomalies after event onset is evident, as perturbation s over the South Pacific maximize from days 2 to 4.

In the cases of "SACZ only" (Fig. 5.8), low height anomalies are located at about 90 W at day -6 from onset, and they gradually shift eastward and strengthen. By day 5, maximum values are at 30 W with magnitudes up to -150 m. The upper-level low anomalies become established to the southeast of the high anomalies over the plains (see Fig. 5.4). This evolution matches the increased magnitude and extent (from the subtropical plains over the Southwest Atlantic) of the precipitation for the "SACZ only" cases. A more localized, intense circulation is shown. The wave activity over the South Pacific appears weaker than in Fig. 5.7 (also seen if Fig. 5.4) but it intensifies over the Indian Ocean. From then on, waves propagate faster towards the east at approximate speeds of 10 m/s. Such faster propagation rates are not observed for the "with SPCZ" cases. An examination of the hydrological processes for the positive cases follows.

This section examines the moisture sources which contribute to the two cases of suppressed SACZ. Different moisture sources, as for cases of enhanced SACZ (Chapter 4), are responsible for differences in precipitation. We start with a description of the low-level field and low-level temperature anomalies. To describe the low-level flow field in relation to the precipitation anomalies, Fig. 5.9 references the 850 mb winds and shows the difference between the two types of positive cases (Fig. 5.9b).

The precipitation anomalies are plotted with the low-level winds in Fig. 5.9a for the "SACZ only" composite. Comparison with OLR anomalies in Fig. 5.2 h shows good correlation between the OLR and precipitation fields. The distribution of positive and negative precipitation (and OLR) anomalies over the Pacific and South America shows characteristics of a warm ENSO signature. Precipitation over the Eastern Pacific is enhanced, and a strong precipitation signal flanks the suppressed SACZ over Northeastern and Southeastern South America (eg. Ropelewski and Halpert, 1996).Low-level circulation features (Fig. 5.9a) include the subtropical high over the South Pacific, a displaced subtropical high over the Southwestern Atlantic whose western periphery flows southward parallel to the Andes, zonal westerly winds at midlatitudes, and a strong easterly flow of tradewinds into the interior of Northern South America.

Differences in wind and precipitation (Fig. 5.9 b) show several features. Both the Atlantic and Pacific subtropical highs are displaced to the southwest. Precipitation is more intense over the SPCZ, over the Pacific ITCZ along 10 N, over the Atlantic ITCZ, and over the subtropical plains of South America for "SACZ only" events. Off the coast of South America, the low-level trough is stronger, as shown with the extension of precipitation from the plains. The "with SPCZ" cases show stronger precipitation anomalies over Northern South America, the Equatorial Pacific and over the region flanking the SPCZ to the east (150 W 20 S). The northerly jet over the subtropical plains is stronger for the "SACZ only" cases (seen with a positive vector in Fig. 5.9b). Over the Equatorial Pacific, the main difference is in the meridional component with a stronger southerly contribution for the "SACZ only" composite. This may be associated with convergence in the area of positive precipitation anomalies along 10 N, (to be further investigated with an examination of moisture flux differences). In the Pacific at about 20 N, the "SACZ only" composite is stronger in the zonal direction in a region where precipitation is more suppressed than in the "with SPCZ" cases.

Precipitation anomalies over the western hemisphere are similar for both cases, with an amplified signal for the "SACZ only" cases. Main circulation differences are found at midlatitudes. This is also apparent in the temperature field (Fig. 5.10). Over the South Pacific and Southwest Atlantic, the "SACZ only" cases have cold anomalies. Over the Atlantic, this corresponds to enhanced precipitation extending from the subtropical plains from the "SACZ only" cases. The presence of these anomalies may suggest a role that cold surges have in locally intense frontal passages over South America. This was attributed earlier to the "SACZ only" composite in reference to the findings of Garreaud and Wallace (1997) (see section 5.2). In Fig. 5.10 a, the temperature anomalies in the Southern Hemisphere are in an opposite phase to those shown in Fig. 4.10 a (for the "with SPCZ" negative cases), particularly over the South Pacific. This further supports the "seesaw" nature of the "with SPCZ" events, as shown in the low-level temperature field. Over South America at about 25S, the temperature differences between the two cases are small, as both show warm temperature anomalies flanking the dry and wet precipitation anomalies. In general, the "with SPCZ" events are warmer than the "SACZ only" events over the western hemisphere.

Temperature anomaly differences are also evident near North America, and this lends to correlations of the events with activity in the winter hemisphere. The "with SPCZ" composite shows significantly warmer temperatures over the Eastern Pacific accompanying positive precipitation anomalies and a strong low-level trough (see Fig. 5.4 d). For the negative cases, a cold anomaly was evident over the Gulf of Mexico for "SACZ only" cases (see Fig. 4.10 c), while for the positive composite, a warm anomaly (the opposite) is shown in the same region near 25 N (Fig. 5.10 c). The position of the low height anomaly (see Fig. 5.4 e) also suggests warm southerly flow over the Gulf. This is verified with Fig. 5.10 e, which shows pronounced anomalous southerly flow at 925 mb. The "with SPCZ" composite shows weak easterly flow in the Gulf of Mexico (Fig. 5.10 d), in contrast to southerly anomalies observed for"with SPCZ" negative events.

The precipitation patterns in the tropics and over South America are similar for the two cases, but as Fig. 5.10 d and e show, low-level flow patterns differ significantly. For the "with SPCZ" cases, precipitation anomalies over the subtropical plains of South America can be tied to the low-level jet and moisture supply from the Amazon Basin. Eltahir and Bras (1993) describe contributions to the hydrologic cycle through a recycling of water vapor from the Amazon. They estimate a 35% recycling ratio. The availability of moisture through evapotranspiration is important to surrounding regions, as illustrated in Fig. 5.10 d. For the "SACZ only" events (Fig. 5.10 e), transport of Amazonian moisture is also important over the plains, but low-level flow from the suppressed SACZ (southward) appears to play an even stronger role in the enhanced anomalous convection. The 925 mb winds turn southward from the suppressed region and approach the plains more directly for the "SACZ only" cases (Fig. 5.10 e) than they do for "with SPCZ" cases (Fig. 5.10 d).

Evaporation and moisture flux are two possible sources of moisture for these cases. Anomalies of evaporation are plotted with precipitation anomalies in Fig. 5.11. For both types of positive composites, po sitive precipitation anomalies are generally associated with negative evaporation anomalies, as less solar radiation is available for solar heat flux. As shown for the negative cases, the anomalies are small. They are at least an order of magnitude smaller than the precipitation anomalies. Thus, the main source of moisture is not from evaporation, but from moisture flux. For the negative cases, negative evaporation anomalies are widespread over the Southeastern Pacific (see Fig. 4.11). Positive evaporation anomalies dominate this region for the positive cases (Fig. 5.11 a,b). A relation of stratus decks to the temperature field is made in section 5.5. However, the role of evaporation in the cloud processes over this region, as assimilated by the reanalysis, is unclear at th is time. An examination of the moisture flux reveals that it is the source of moisture for these events.

The moisture flux is the major source of moisture, similar to the conclusions in Chapter 4. Fig. 5.12 a and b show the moisture flux for both types of positive cases, with values less than 75 g/(cm*s) not plotted. For both cases, convergence of moisture flux is pronounced in regions corresponding to positive precipitation anomalies. An example of this is over the Pacific at 10 N in Fig. 5.12 b, where strong precipitation anomalies are also located. Over the subtropical plains of South America, north-to-south vectors parallel the Andes. They correspond with moisture transported by the low-level jet that sets up as the Atlantic subtropical high shifts westward. Moisture flux convergence is weak over the SACZ where convection is suppressed. For the "SACZ only" cases, significant convergence of moisture flux extends from the South American continent well into the Atlantic. Over the Western Atlantic basin, the eastward flowing branch of the subtropical high brings in Atlantic air which is channeled southward by the Andes.

The differences in moisture flux show the nature of the anomalies for the two types of cases. Differences are significant, as maximum arrows are greater than 50 percent of the original values. (Fig. 5.12 c). Values of moisture flux less than 25 g/(cm*s) are not plotted. Shaded values show the convergence of moisture flux for the "with SPCZ" cases for use as a reference. The flux of Atlantic air into the continent is most pronounced for the "with SPCZ" composite. Over midlatitudes, the baroclinic structure is evident in the wave-like orientation of the vectors. Other general features include the stronger convergence of moisture flux over the Central Pacific at both 10 N and 22 N from the "SACZ only" cases (where precipitation shifts northward, as it also does for the negative cases). Along 5 to 10 N, the meridional component of moisture flux clearly sustains the convection in the form of stronger southerlies. Also, flux convergence over the subtropical plains of South America into the Atlantic is stronger for the "SACZ only" composite. This agrees with the heavier precipitation found over this region for these type of events. In addition, the flux of warm, moist air from tropical latitudes is stronger over the South America for the "SACZ only" cases (as shown by the positive vector in the differences). This is also consistent with the warm anomalies located northward of the precipitation anomaly band (Fig. 5.10). The results found with the moisture flux show that the origin of moisture for the precipitation anomalies, as concluded for the negative composites, is in the flux of moisture. The next section compares assimilated OLR and precipitation to observational (ISCCP) data for the cases of suppressed SACZ.

The composites of positive cases used for comparison to the ISCCP and observed OLR (Liebmann and Smith 1996) dataset are for the years 1985 to 1991. This provides a reasonable way to validate some of the conclusions made with the ECMWF dataset. Composites of OLR and precipitation are compared to observations of OLR and high, middle and low clouds produced by ISCCP. A limitation in this comparison is ISCCP data is missing from certain critical areas, such as over the Pacific Ocean from about 150W to 130W.

Precipitation tends to correspond well with the OLR over tropical and subtropical latitudes in Fig. 5.13. The precipitation and OLR show general features of the "with SPCZ" events, such as suppressed convection over the SACZ and precipitation signals over the SPCZ and the ITCZs in both the Pacific and Atlantic. Deviations from observed OLR (Fig. 5.13c) suggest that the ECMWF produces too few clouds (up to 40 w/m2 difference) over South America and over the Pacific when compared to observations. The assimilated precipitation and OLR match well with the middle and high cloudiness in the ISCCP dataset (Fig. 5.14). This good correlation can be attributed to the prognostic equation for cloud water used in the ECMWF reanalysis.

Differences in OLR and precipitation are taken in Fig. 5.15 a and b. The precipitation and OLR fields correlate well and show positive and negative precipitation anomalies which have been discussed in previous sections. In Fig. 5.15 c, the differences in the observed OLR are taken. Comparison of Fig. 5.15 b with c shows that assimlated differences generally match observations. For example, over the Central Pacific in the Northern Hemisphere, OLR values are lower for the "SACZ only" cases in both analyses. This is also the case over the SPCZ, where convection is more intense for "SACZ only" events, though differences are higher for the observed fields. ISCCP differences are shown in Fig. 5.16. At upper levels, the mid and upper-level cloudiness over the Pacific tend to agree with assimilated results, though much of the area is void of data. Over South America from 5 to 10 S, the "with SPCZ" cases produce more precipitation in the assimilated reanalysis (Fig. 5.15 a), and this agrees with the higher amounts of mid and upper-level cloudiness shown in the ISCCP differences (Fig. 5.16 a and b).

Stratus decks over the Southeastern Pacific are significant features shown for these events in the ISCCP low cloud analyses. One explanation of the magnitude and extent of the clouds may be a relationship to the temperature field. At low levels, diffe rences in stratus over the cool waters of the Southeastern Pacific (Fig. 5.16 c) are similar to the results found for the negative cases, as the "with SPCZ" cases have more stratus off the coast of South America. The amount of low cloudiness, though, is lower for positive cases, as shown in the differences in Fig. 5.17 a. Negative events have 5 to 10 percent more cloudiness over the Southeas tern Pacific. Warmer temperatures in this region for the negative "with SPCZ" cases may account for the higher stratus amounts, as evident with 850 mb reanalysis temperature differences (Fig. 5.17 c). Low cloudiness differences for "SACZ only" events (F ig. 5.17b) show higher amounts of stratus for positive events. This corresponds to warmer temperatures off the coast of South America (Fig. 5.17 d). Overall, comparisons of observed OLR and cloudiness to ECMWF OLR, precipitation and temperature fields suggest a reasonable depiction of observations.

Ch.6 Discussion and Summary

The South American continent exhibits a unique blend of steep mountains, tropical forests, desert areas and fertile subtropical plains. The Bolivian altiplano provides an elevated source of sensible heat during summer (e.g. Rao and Ergodan 1989). It influences large-scale vertical motion patterns and results in atmospheric overturnings and associated rainfall and convective heating. A high pressure system is found during summer at upper levels over this 3000 m-high plateau. Nevertheless, the summer circulation over South America has been qualitatively reproduced as the response to transient tropical convection (Silva Dias et al. 1983) using a linear model similar to that of Gill (1980). Resulting features in this model include the Bolivian high, suggesting it may be more a result of tropical convection than the elevated heat source. A prominent feature of the summer convection over South America is the South Atlantic Convergence Zone (SACZ), a convective band which extends, at times, from the Amazon basin into the Atlantic Ocean. The present study focuses on the intraseasonal variability of the SACZ and describes its horizontal structure.

The 30 to 60 day oscillation (e.g., Madden and Julian 1971) is a strong modulator of tropical convection on the intraseasonal time scale. It was originally identified as a pulse of organized convection that propagates eastward along the equator from the Indian Ocean to the Central Pacific. It has been shown to propagate all the way around the equator and to be strongest during austral summer. Strengthening of the SPCZ and the subsequent enhancement of the SACZ have been documented by NPM97, who also show the existence of a meridional seesaw of dry and wet conditions over tropical and subtropical South America. The present study further examines the intraseasonal variation of precipitation regimes over South America. It seeks to identify the importance of the upstream intraseasonal modulation of the SPCZ on the SACZ and related meridional propagation patterns.

This is accomplished using rotated empirical orthogonal function analyses of filtered OLR anomalies confined to two regions: one includes the SPCZ from the dateline eastward to 20E; and the other extends from 100 W to 0 W, including South America and the SACZ. Four convective patterns are identified in this fashion: those with an enhanced/suppressed SACZ with a well-defined SPCZ (referred to as "with SPCZ"), and those with an enhanced/suppressed SACZ without a clear signature in the SPCZ ("SACZ only"). The investigation describes midlatitude patterns, synoptic transients, and moisture sources for all four of these events. Main results follow:

1) An enhanced SACZ is associated with a suppressed SPCZ (as in NPM97) and is preceded about 8 days prior by a pronounced SPCZ. These are features of the intraseasonal oscillation. The seesaw pattern between the SACZ and precipitation over the South American plains is also apparent in these analyses. Both an enhanced and a suppressed SACZ with a well-defined SPCZ are more frequent during warm ENSO events. This result is consistent with stronger convection in the Eastern Pacific, a typical signature of ENSO events.

2) Four well-defined, northwest-to-southeast-oriented precipitation bands located over surface low pressure are found for "with SPCZ" cases. These are located from the SACZ eastward, east of Africa, over Australia, and in the Central Pacific, with extensions from the tropics into midlatitudes. These bands acquire a more zonally-oriented tilt as they merge into the westerly flow. Approximately 5 waves are found in the 200 mb height anomaly pattern between 40 S and 50S, suggesting a possible link between precipitation and height fields. This is not the case for the "SACZ only" events. They show strong precipitation anomalies over South America and the dateline in the Southern Hemisphere subtropics. A dominant wave number 2 pattern in midlatitudes may be related to these two subtropical convective anomalies.

3) The wave number 5 pattern in the height field composite, referred to above, is found in conjunction with high synoptic variability at midlatitudes over the Southern Hemisphere. This high wave number pattern is consistent with increased synoptic wave activity and possibly increased frequency of frontal passages over South America and the neighbouring Atlantic Ocean. "SACZ only" cases appear less variant with steady in-situ amplification and no apparent meridional link between the tropics and the subtropics. Both types of negative events consist of a similar number of cases (14 "with SPCZ" and 16 "SACZ only"), suggesting an equal role for both types in the overall climatology of the SACZ.

4) The larger precipitation amounts over tropical South America for "SACZ only" cases, compared to "with SPCZ" events, are sustained by a stronger convergence of moisture flux. Atlantic air is advected into the continent by the northerly branch of the subtropical high. This is the largest contributor to moisture flux for the SACZ. Minor contributions are due to the northwesterly flow channeled by the Bolivian altiplano, which is relatively stronger for the "with SPCZ" cases. These analyses show no contributions to the SACZ from southerlies.

5) A number of similarities exist between cases of enhanced and suppressed SACZ. First, evolution of OLR anomaly composites are similar both in positive and negative "with SPCZ" events in that the seesaw pattern is apparent (cases with suppressed SACZ activity are denoted as positive cases). No convective precursors are apparent for the "SACZ only" composites, where higher precipitation values are more evident over South America. Second, "with SPCZ" cases exhibit more synoptic activity over the South Pacific Ocean. Third, they are more frequently found during ENSO warm events. Fourth, they are associated with a larger number of waves in Southern Hemisphere midlatitudes. Therefore, the chosen stratification (with and without a well-defined SPCZ) has successfully identified two different patterns of intraseasonal variations of summer rains over South America.

6) Positives events also coincide with an enhanced SPCZ and ITCZ. The latter is displaced about 5 to 10 degrees north. The SPCZ and ITCZ patterns are reminiscent of those associated with the mid-summer climatology.

7) Large differences are apparent in the moisture flux between the positive and negative cases. A low-level northerly jet for positive cases is responsible for transporting moisture from the Amazon Basin. The eastward-flowing branch of the Atlantic subtropical high is thus deflected southward. The flux of Atlantic air into the continent is most pronounced for "with SPCZ" composites.

Comparisons of ECMWF reanalysis products with ISCCP cloud coverage indicates that the analyses do a credible job in linking regions of convective precipitation with mid- and high-level clouds. This is expected from the ECMWF physical parameterizations, as they carry prognostic equations for the cloud/ice content. OLR fields from the analyses are found to be overestimates of those derived from satellite observations, with differences peaking at 40 w/m2 over tropical convection. The analyzed evaporation anomalies appear negatively correlated with cloud cover, indicating no local feedback between surface procesess and precipitation anomalies.

The ECMWF analyses resolve well-known features of the seasonally averaged climatology, including patterns of precipitation and evaporation. They also correctly resolve the convergence zones of the Southern Hemisphere; and over South America, they resolve the seasonal evolution of the Bolivian high.